27

The physical system design of the Atari 800 was pretty radical, when compared to the 3 other canonical home computer designs of the late-1970s: TRS-80 Model 1, PET, and Apple ][.

As the photo of the Atari 800 main and CPU board shows, the main board was slotted, but not for the purpose of interfacing general peripherals and enhancements, like in the Apple ][ or later IBM PC. Rather, the Atari 800 had a very modular construction for which not only the CPU, but also all RAM, ROM, and firmware were added via cards. Of course, the 1977 trio were all built with an integrated motherboard that included CPU, RAM, and ROM instead of Atari's modular approach.

What would be the rationale for the Atari designers to take this radically different approach compared to the aforementioned competitor systems?

(Backplanes weren't new, but the Atari 800 came out two years after other home computers had abandoned such a design. see above)

Atari Boards

Improve this question
6
  • 8
    It's not that unusual of a design. The Altair 8800, the first successful personal computer, was like that. It also how most minicomputers were designed.
    – user722
    Commented Apr 30, 2017 at 15:29
  • 3
    Ok. But the Atari 800 was neither a niche hobbyist kit (like the Altair) nor a minicomputer. If it wasn't "new ideas", what rationale would lead the Atari designers to adopt "old ideas".
    – Brian H
    Commented Apr 30, 2017 at 15:32
  • 3
    Basically all early Micro- and Minicomputers which had CPUs made from discrete elements used this backplane design. You'd have one or several boards for the CPU, one or several for core memory, and the rest for peripherals. So it's rather the Apple etc. which used a different approach, providing slots only for peripherals. Of course this was only possible because the CPU was now on a single chip, and RAM chips had enough capacity to put them on the motherboard.
    – dirkt
    Commented Apr 30, 2017 at 16:18
  • 1
    @dirkt If you are implying that Atari designers were Luddites, that could be an interesting perspective for an answer.
    – Brian H
    Commented Apr 30, 2017 at 16:46
  • 2
    If course they were not Luddites. But the point is that while you make a backplane design seem totally radical and new, in fact it wasn't: It was really really well know at the time, everyone who had used a Micro or Mini was familiar with it, and it's a lot more flexible. So it was completely natural to start out with a backplane design - one always could integrate it later, just like it happened.
    – dirkt
    Commented Apr 30, 2017 at 17:05

4 Answers 4

Reset to default
35

The physical design of the Atari 800 was radical, but not for reasons related to the aspects you highlight.

As others have commented, using multiple boards for a system was pretty standard at the time. S-100 systems for example were based on a backplane, with system features implemented on multiple boards connected to that. (The Atari 400 and 800 were designed using a Cromemco S-100 system — I don’t know whether that had any direct impact though.) The Atari 800 has three main boards: the “motherboard” (containing POKEY, PIA, the memory decoder, and all the system connectors), the CPU board (containing the 6502, CTIA or GTIA, and ANTIC), and the power supply board. User-accessible boards (initially sold as cartridges) provide the ROM (“system personality”) and RAM (not firmware though — what we’d consider firmware nowadays was provided by the external devices themselves).

The main drivers behind the design appear to have been:

  • user-friendliness, which led to building an enclosed system with well-protected user-accessible parts, such that the user could upgrade the system without being exposed to such things as the power-supply;
  • compliance with FCC requirements, in particular because of the inclusion of an RF modulator — this is the reason why the system is so heavily shielded.

Both of these were radical departures from the other personal computers of the time; the first in particular had far-reaching consequences. Instead of opening up the whole system to access expansion cards, Atari 800 users could open the first flap to connect cartridges containing software, the second flap to change their system’s ROM or RAM, and anything else was connected to the joystick or SIO ports. To keep all this safe for users, the flaps had interlocks which ensured that the system was powered down when the flaps were opened; unless the users went out of their way to circumvent them (well, it wasn’t that hard really, but it couldn’t happen by accident), there was no risk of connecting or disconnecting a cartridge while the system was powered up. To keep all this easy to use, the designers had to come up with some rather novel ideas: the SIO bus in particular was far ahead of its time, with device-supplied drivers loaded automatically (the descendant of SIO is USB, amazingly enough).

Coming back to the multi-board design, it does have some nice features, which might explain why it was kept at least for the initial system. Having a separate power supply board means that the main board and power supply board can be designed separately, by experts in the respective fields — these different kinds of boards require different kinds of expertise to design well. It also means that the main board can be bread-board designed without too much regard for the power supply; Goldberg and Vendel’s Business is Fun shows that bread-boards were used extensively. Keeping the CPU, TIA and ANTIC on a separate board might have been relevant because that whole section of the computers took a while to settle; the same book shows that the designers were still contemplating different CPUs late in the system’s development.

One very nice benefit of using different boards like this, is that the CPU and RAM boards could be re-used as-is for the Atari 400 (which was re-designed, or perhaps just re-specified and designed, in a hurry not long before launch). The Atari 400 shares the same design, with a separate power supply board, main board (which includes the ROMs), and separate CPU and RAM boards (neither of which are supposed to be user-serviceable in the 400).

Another nice benefit is that the base memory configuration could easily be changed to take advantage of price variations in the market. The Atari 800 was supposed to ship with 8K of RAM (hence its name — the Atari 400 was planned with 4K), but as RAM prices fell, its base configuration improved progressively until it ended up shipping as standard with 48K a few years later. This would be possible with socketed memory chips on a big motherboard, but that approach is a little more restrictive. The Atari RAM boards are in some ways like early SIMMs, and allowed Atari to increase chip density without changing the motherboard (the boards were available in 8K and 16K variants, with the same number of chips).

(The design wasn’t without its flaws of course. The initial cartridge design used for ROM and RAM made for poor airflow and created thermal problems, so the metal shielding around the cartridges was removed in later models, and the second flap’s plastic latches were replaced by screws. For completeness’ sake it’s worth noting that expansion wasn’t limited to the external ports entirely; the last RAM slot could be used for other kinds of expansion boards.)

The Atari Museum’s page on the Atari 800, the Wikipedia entry, and the above-mentioned Atari Inc.: Business is Fun all give further background on the design of these systems.

Improve this answer
9
  • 5
    The Atari 800 was an exceptionally well built and underrated computer for its day. Considering the time frame it was released, the fact it had hardware sprites, hardware scrolling, amazing audio and even S-Video support was amazing. Not to mention that gigantic metal RF shield inside it! During a nuclear war, the only things left alive would be roaches and Atari 800's. (NOT XL).
    – cbmeeks
    Commented May 1, 2017 at 13:55
  • 2
    As to system specs, I think it's a case of necessity being the mother of invention: the version I read had the GTIA and POKEY queued up and ready to go as successors to the 2600's TIA, CPU-powered kernels and all, but the rest of the industry's move to video output that doesn't need its hand constantly held forced Miner and co to draw up the ANTIC. So sprites, etc, carry over but you also get DMA provision. Which leads directly to the Amiga and therefore indirectly to the Lynx and 3DO. It's a shame Atari weren't more aggressive on pricing and marketing.
    – Tommy
    Commented May 1, 2017 at 17:12
  • 2
    @cbmeeks: The sprites were an offshoot of the Atari 2600's sprites, and were intended to reduce the amount of memory required for a usable game. The Atari 2600 has 128 bytes total RAM, but can display 160 by 200 pixels using 128 colors. The Fairchild Channel F had more than sixteen times as much RAM, but could only show about a quarter as many pixels and was limited to four colors.
    – supercat
    Commented May 1, 2017 at 18:25
  • 1
    @supercat thanks, but I was just mentioning how I thought the Atari 800 was a ground-breaking computer for the year it was introduced. Especially when you factor in what it could do.
    – cbmeeks
    Commented May 1, 2017 at 18:39
  • 1
    @cbmeeks: According to Joe Decuir (one of the designers of the Atari) the cartridge slot power interlock was included for FCC compliance. When the Atari was introduced, consumer devices that used high-frequency signals used them in relatively isolated places that could be easily shielded very well. The people who had written the FCC rules had thus mandated very strict RF emissions limits because complying with those limits wouldn't add much to the cost of products or limit what they could do. FCC rules for computers were loosened soon after that, but a lot of design...
    – supercat
    Commented Jul 2, 2017 at 19:41
3

I'm pretty sure the main reason for the slots was to keep the cards vertical, for cooling purposes. Airflow went straight up the slots and out the top.

In theory, they could have implemented this with two PCBs, and stacked them. This would trap significant heat trapped between the two. Recall that they had to take the packaging off the cards even in the 800 because of heat problems.

They could have put everything on the verticals, but you'd still need a backplane anyway, so you may as well use it for something other than just power.

Improve this answer
3

I know this is a bit of an old thread, but just wanted to note that the design also allowed for expedited repairs should a consumer so request - the defective board would be swapped (or the motherboard if it were the culprit) and the unit returned rather rapidly. The defective board would then be refurbed for a future repair.

Improve this answer
1

The 400/800 design had some rather late design changes over the FCC.

My demo model the summer in high school selling them was an extremely early production model. Serial number 49, iirc. It had the earlier graphics chip with a larger color palate in some modes.

Anyway, the massive metal pot was not originally intended, nor was the all-serial model of communication.

There is an SS-50 edge connector on the back end of the motherboard, which was supposedly meant for peripherals.

One day I suggested, and the partner present agreed, and we disassembled it and sawed off enough metal to access the edge connector. The bright idea was faster peripherals.

The other partner had a bit more business sense, and it stopped there.

This then had the consequence of those silly serial drives. "Slow" was not an adequate word for their performance. Not only was the maximum data rate slow, but they were of the "CPU in every design" school of thought. And they took long enough to process the reading of a sector that the start of the next sector had passed the heads by the time the controller was ready.

They later came out with the "rev C rom" drive, which could be retrofitted to older units in the field, which skipped sectors and seriously increased speed (well, when reading data also written with one).

The 400/800 design was designed for the regulatory environment that they anticipated or assumed would be in place at the tine of the release. It had to be modified to deal with the rules that were actually in place.

I don't recall a single computer with both an RF modulator and an external parallel bus. For that matter, I can't even think of any that had both the modulator and a centronics printer port, although there may have been.

Improve this answer

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .